“…Seven of the 72 included studies [ 56 , 59 , 60 , 62 , 71 , 80 , 87 ] were scored at a high overall risk of bias (eTable 3); all investigated opioid-related adverse events. More than one-third of included studies (25 studies) [ 35 , 36 , 45 , 47 , 51 , 53 , 56 , 57 , 59 , 62 , 63 , 67 , 71 , 73 , 75 , 77 , 80 , 85 , 87 , 88 , 90 , 92 , 94 , 95 , 101 ] were considered as having high or unsure risk of attribution bias due to the lack of reporting of loss to follow-up in randomised controlled trials or data completion rate not reported in observational studies.…”